Sociodemographic, health and fracture proles of a 4-year cohort of 266,324 rst incident upper extremity fractures in Ontario.

The purpose of this study was to describe 1st incident fractures of the upper extremity in terms of fracture characteristics, demographics, social deprivation and comorbid health proles. Methods: Cases with a 1st adult upper extremity fracture from the years 2013 to 2017 were extracted from administrative data in Ontario, (population 14.3M). Fracture locations (ICD-10 codes) and associated characteristics (open/closed, associated hospitalization within 1-day, associated nerve or tendon injury) were described by fracture type, age category and sex. Fracture comorbidity characteristics were described in terms of the prevalence of diabetes, rheumatoid arthritis; and the Charlson Comorbidity Index. Social marginalization was expressed using the Ontario Marginalization Index (ON-Marg) for material deprivation, dependency, residential instability, ethnic concentration. The highest of fractures: distal (DRF, metacarpal phalangeal distal phalangeal proximal humerus clavicle radial and scaphoid The most prevalent multiple fractures included: multiple and ulna fractures fractures occurring in multiple regions of the upper extremity (8.7K), or multiple regions in the forearm Fractures most common in – 40-year-old men included metacarpal and nger A large increase in fractures in women the age of occurred for: DRF, PHF and radial Tendon 8.2% multiple nger fractures) nerve injuries open in 4.7%, highest for distal phalanx (23%). Diabetes occurred in 15.3%, highest in PHF (29.7%). Rheumatoid arthritis occurred more commonly in women (2.8% vs 0.8% men). The Charlson Index indicated low comorbidity (mean=0.2; median=0: 2.4% 3+), highest in PHF (median=0; 6.6% 3+). Higher fracture burden was related to instability (excess of fractures in lower 2 quartiles 4.8%), although social indices varied by fracture type. Fracture specic prevention strategies should consider fracture-specic age-sex interactions, health, behavioural and social risks the 50–65 of inection for women was observed for scaphoid fractures, radial head fractures, olecranon fractures, multiple fractures of affect fractures through multiple mechanisms including sex differences in genetic risks that affect the size and quality of bone, gender-based differences in behaviour norms that affect behaviours and life roles, and social differences that affect work and life exposures. African American and Hispanic men have been reported as having higher incidence and younger onset of hip fractures than white men[33], which is consistent with what we found for some upper extremity fractures, although not the classic fragility-subtypes.

1. Injury patterns: fracture numbers, fracture subtypes (location, % open), associated nerve or tendon injuries, injury seasonal patterns and hospitalization rates 2. Comorbid health status: overall comorbidity (Charlson comorbidity, presence of diabetes, rheumatoid arthritis) 3. Sociodemographic pro les (age, sex/gender, income quintile, rural/urban status, material deprivation, dependency, residential instability, and ethnic concentration). Patient data were extracted from relevant administrative data sources that arise from hospital admission/discharge (Discharge Abstract Database), outpatient surgery (Same Day Surgery), emergency visits (National Ambulatory Care Reporting System), physician billing data (Ontario Health Insurance Program [OHIP] database), and general use data sets containing sociodemographic data (Registered Persons Database). ICES-developed disease cohorts were used to identify rheumatoid arthritis (Ontario Rheumatoid Arthritis Database) and diabetes (Ontario Diabetes Database). Data were linked to associated demographic and fracture information using unique, encoded identi ers, and analyzed at ICES Western. Data were depersonalized or aggregated for con dentiality reasons, and precautions were taken to ensure indirect identi cation was not possible (e.g. suppressing small cells).

Methods
Fracture Identi cation/Description. This retrospective cohort study includes all adult patients who visited the emergency room with an upper extremity fracture from January 1, 2013 to December 31. 2017. Only the rst such event was eligible for inclusion. To ensure all fracture-related diagnoses were captured, we also included fractures reported during ER visits and hospital admissions within 5 days of the initial ER visit. Fractures coded as 'questionable' were counted if there was other supporting evidence of fracture from an ER visit, hospitalization, or OHIP record. In addition to data cleaning exclusions, we excluded non-Ontario residents and those with evidence of an upper extremity fracture, as an adult, within the previous 10-years (previous pediatric fractures were permitted). To help make sure we identi ed all previous fractures, we also excluded individuals who were not eligible for healthcare services in Ontario for this full 10-year look-back period.
Fracture pro les Nerve injury was grouped as a major nerve injury when codes for speci c major nerves in the upper arm (ulnar, median, radial, axillary, musculocutaneous), forearm (median, ulnar, radial), or wrist (median/ulnar) were identi ed; and as any nerve injury by collapsing all speci c and non-speci c nerve codes for each of these 3 areas. Hospitalization within 1 day was extracted.

Patient Pro les
A descriptive pro le of the entire cohort and major fracture subtypes was constructed to describe demographics, fracture characteristics, comorbidity and social marginalization; considering age and gender pro les for speci c fractures. Age was classi ed based on clinically relevant subgroups: aged 18 to 40 (young adults), 41 to 50 (younger middle-age adults), 51 to 65 (older middle-age), 66 to 80 (older adults), 81+ (very old adults). The age groups were selected to represent anticipated different health and fracture risk pro les, with young adults having the best bone health and being more likely to engage in higher fracture risk activities. Younger middle-aged adults were considered as being transitional in terms of activity and bone health, while older middle-age individuals (51-65 years) are those with emerging fracture risks. Older adults were considered more likely to have associated comorbidities and the very old to have a higher probability of frailty.
Diabetes (from ODD) and Rheumatoid Arthritis (from ORAD) diagnoses were identi ed using ICES de ned cohorts. People with diabetes were identi ed as all individuals in the fracture cohort who appeared in the ODD before the fracture index date, based on having at least one hospitalization or two physicians' service claims for diabetes within a 2-year period. The ODD has high sensitivity (86%) and speci city (97%) for identifying diabetes when compared to extraction from primary care charts [17]. The RA cohort algorithm in ORAD has high sensitivity and speci city compared to a rheumatologist diagnosis [18,19].
We attempted to extract data on osteoporosis. However, since a 1st upper extremity fracture at the age of 50 or older may initiate the diagnostic process, Bone Mineral Density data was rarely available. Since this data is used in de nition of osteoporosis, we suspected underreporting. [20] Similar problems arose in extracting an osteoarthritis diagnosis, since this is not consistently coded unless it is the primary reason for the visit. Given validity concerns with these 2 diagnoses, we did not report osteoporosis or osteoarthritis comorbidity, despite their importance.
Overall comorbidity indicators were derived from the Charlson comorbidity index [21]. The Charlson Comorbidity Index categorizes comorbidity using International Classi cation of Diseases (ICD) diagnosis codes. Each comorbidity category has an associated weight (from 1 to 6), based on the adjusted risk of mortality, and the sum of all the weights results in a single comorbidity score for a patient [21,22]. A score of zero indicates that no comorbidities were found. The Charlson Index is the most extensively studied comorbidity index for predicting mortality [23].
The Ontario Marginalization Index (ON-Marg) was used to assess social inequities (material deprivation, dependency, residential instability, ethnic concentration) [24]. ON-Marg was developed using a theoretical framework based on previous work on deprivation and marginalization and empirically derived using principal component factor analysis. A quintile score is associated with geographic areas, so that each quintile represents 20% of the reference population. To describe marginalization, we compared the top two quintile values to the bottom 2 quintiles. We graphed the size and direction of any excess of fractures in the lower double-quintile (expected 40%). For the ON-Marg Index, those in lower quartiles are considered more deprived or more marginalized.
Maternal deprivation re ects socioeconomic status, education and single-parent families and is comprised of indices including the proportion without a high school diploma, single-parent families, proportion of income from government support, proportion of those over 15 who are unemployed, proportion classi ed as low income and the proportion living in houses classi ed as requiring major repair. Ethnic concentration re ects the proportion of the neighborhoods who are recent immigrants or visible minorities. Dependency re ects the extent to which the population is not supported by employment income and includes indicators of proportion of people over 65, a ratio of younger to older people, and the proportion not participating in the workforce. Residential instability re ects home security, ownership and occupancy and is comprised of indicators of living alone, proportion of adults, number of people per dwelling, proportion who are single/divorced/widowed, and the proportion who have moved in the past 5 years. Marginalization and co-morbidity have previously been associated using ICES data [25].

Analysis and hypotheses
Our purposes were descriptive and the analysis focused on counts or percentages that described the number of fractures by type and disaggregated by gender and age groups; describing pro les in terms of the demographics, fracture characteristics, comorbidity and social indicators. We expected that young adults would be most susceptible to higher energy "misfortune" fractures; whereas the older middle-aged cohort would contain both active and compromised individuals. We expected gender differences in young adults to re ect higher risk-taking or involvement in fracture susceptible activities in young men. We expected sex differences in bone health to emerge by the 50-year-old age category and to be more pronounced in fractures that are considered fragility fractures such as DRF and PHF. We expected that frailty fractures would be prevalent in the old and very old subgroups. We considered differences in social deprivation of 5% between upper and lower categories to be potentially relevant and 10% or more very likely to be relevant.

Results
After exclusions from the potential cohort (n = 506,071), mostly related to pediatric fractures (n = 174,378), we were left with a cohort of 266,324 people with a rst upper extremity fracture occurring over a four-year interval (n = 44,236 were excluded because they had prior fractures) (Fig. 1). Overall, looking grossly at the fracture distributions by region the most common fracture sites were the hand (94K), followed by the wrist (80K), shoulder (49K) and elbow (35K) - Fig Table 1. Uncommon fractures included coronoid (n = 327), isolated radius shaft (n = 216), isolated ulnar shaft (n = 170) and Monteggia (n = 25). The mean age of the cohort was 51.5 years and was slightly female dominant (51.5%) ( Table 1). When examining the different demographic pro les of major fracture subtypes, PHF demonstrated the oldest mean age (68.5) and metacarpal fractures the youngest age (37.9) ( Table 1 Table 1). The fractures that were most male predominant were distal phalangeal fractures (71.1%) and metacarpal fractures (70.5%) ( Table 1).
Examining age groups for men versus women for speci c fractures provided further insight that sex/gender modi es the age patterns for speci c fractures. For DRF (Fig. 15), there is a large excess in terms of women presenting with a rst DRF starting with the 51-65-year-old age category and continuing at older ages. For metacarpal and phalanx fractures (Figs. 17 and 18) there is a substantial male excess of fractures in the 18 to 40-year-olds, which is less substantial in the middle-aged categories and reverses so that there is a small excess of women after the age of 66. For the distal phalanx (Fig. 19), a male predominance is present in all age groups, although this is most predominant in the 18 to 40-year-olds. In the oldest group, few DP fractures were recorded, but there is a small excess of women. For PHF (Fig. 11), a small number of fractures and a small excess in males present in the 18 to 40-year-old group; thereafter, the excess is in women with a marked increase in the numbers of fractures and the female to male differential starting at the 51 to 65-year-old age group. For clavicle fractures (Fig. 10), many fractures occur in the 18 to 40-year-old group with a substantial excess in males. An excess in men occurs throughout the lifespan until the 50-65-year-old group. Thereafter in the 66-80-year-old age group the numbers of clavicle fractures become equivalent across men/women and at 81+, there is a greater number of women presenting with clavicle fractures. For radial head fractures (Fig. 13) there was an excess of men presenting in the 18 to 40-year-old group, thereafter more women were present in each age cohort; with the largest gender differential being observed in the 51-65-year-olds. For scaphoid fractures (Fig. 16), multiple fractures of the radius and ulna (Fig. 22), and humeral shaft fractures (Fig. 12) men were more prevalent in the youngest group  and thereafter women became more prevalent. More men than women presented with multiple shoulder fractures (Fig. 21), fractures in multiple regions of the UE (Fig. 20) and olecranon fractures (Fig. 14) in the two younger age groups; but thereafter women were more predominant. Men exceed women with respect to scapular fractures ( Fig. 9) for all but the oldest age group.
Most of the patients were classi ed as coming from urban settings (81.9%; rural 16.2%) Table 1. The only fracture where the proportion of rurality exceeded 20% was for distal phalangeal fractures (21.1% rural).
Overall, only 4.7% of fractures were coded as open fractures; and nerve (0.3%) and tendon (0.6%) injury was rarely reported (Table 2). A higher proportion of nerve injuries was recorded for fractures in multiple regions (1.1%) and phalangeal fractures (0.5%). The fracture type that was most commonly recorded as an open fracture was a distal phalangeal fracture (23%). In total 7.6% of the fractures were associated with a hospital admission. Hospital admission rates were highest for fractures occurring in multiple upper extremity regions (37.2%) and olecranon fractures (26.2%). Overall, the summer season had the highest proportion of fractures (27.3%) ( Table 2). Fractures that occurred most commonly in the winter (> 28% occurring between January and March) were DRF, PHF, and olecranon fractures. Fractures that had a higher predominance in the summer (> 28% occurring between July and September) included: clavicle, radial head, fractures in multiple regions of the upper extremity, and distal phalangeal fractures.
Overall, the cohort of patients with a 1st upper extremity fracture represents a relatively healthy cohort as re ected in Charlson Index scores where 91.8% scored a grade 0 (mean 0.2) ( Table 3). The proportion of cases with diabetes was 15.3% and RA was 1.8%. Higher proportions of diabetes were seen in those with PHF (29.7%) and DRF (19.9%); or approximately 1 in 3 PHF and 1 in 5 DRF. Higher comorbidity on the Charlson index was re ected in higher mean scores for PHF (0.5), and higher proportions of individual who scored 3+: PHF (6.6%), clavicle (3.4%), multiple upper extremity regions (3.1%) and olecranon fractures (3%). The percentage of patients with diabetes for different fracture types is reported separately for men and women (Fig. 23). In PHF, DRF and olecranon fractures the percentage of patients with diabetes were higher for men; whereas for all other fractures the percentage with diabetes was higher for women. For all fracture types a greater percentage of women had RA (Fig. 23).  When marginalization was examined as a variance from the expected 20% in the most marginalized subgroup (lowest quintile) separately for men and women, some additional patterns emerged. Income, ethnicity and deprivation continued to demonstrate little impact on fracture prevalence for both men and women (Fig. 24). However, dependency had a large negative impact on fracture volumes for PHF on women (32.8% versus expected 20%), and this excess was present for men, but to lesser extent. For DRF and clavicle fractures, dependency was associated with an excess of fractures in women, but had little impact on men. Living in the most unstable neighbourhood was associated with an excess in most fractures' types, with one exception being DP fractures in men where a slightly lower than expected proportion occurred in the lowest quintile (18.1% versus expected 20%). See Figs. 24 and 25. Instability had more impact on women for DRF, clavicle, and phalangeal fractures; whereas, the excess in the lowest quintile was slightly higher for men in PHF and RH.

Discussion
This paper provides new information on the fracture, health, and social marginalization pro les of a large cohort of 1st incident upper extremity fractures.
Overall, the hand was the region most affected, re ecting high volumes of nger fractures. The overall ratio of women to men being 51.5% versus 48.5% masked dramatic differences in age-sex pro les that varied substantially across different fracture types; reinforcing the importance of considering sex and lifespan interactions. We selected age categories that re ected different bone health stages of the lifespan and found a variety of sex-lifespan fracture patterns. While it is commonly considered that DRF consists of a bi-modal distribution of young men and older women, we found that women were more prevalent in all age cohorts; although the differences were small in the 18-40-year-old age group. Conversely, PHF which are commonly considered a fragility fracture do have a small excess of men in the 18-40-year-old group; and the large excess of women is not evident until the 50-65 years subgroup. This pattern of in ection for women at 50 years of age was also observed for scaphoid fractures, radial head fractures, olecranon fractures, multiple fractures of the shoulder or multiple fracture of the entire upper extremity. This reversal of sex predominance between the young and older subgroups is thought to indicate different mechanisms of fracture, with younger people being likely to incur "misfortune" fractures that occur due to higher energy trauma, while fractures in older people are more likely to occur with low trauma when bone quality is compromised. Our data raises questions about which fracture subtypes beyond the commonly presumed fragility fractures (DRF and PHF) should be considered as potential indicators of compromised bone health, since similar in ections were observed across some fractures types not typically considered fragility fractures.
Some fractures including scapula, metacarpals, nger fractures and distal phalanx have a high preponderance in young men. These pro les may re ect differences in misfortune fractures which can arise due to greater participation in fracture-risk activities/behaviours e.g. contact sports, greater risk-taking, ghts/aggressive actions, manual occupations, outdoor work etc. Gender differences in engagement in these exposures could explain the higher predominance of some fractures for young men. For example, metacarpal fractures have a high predominance amongst young men, but this differential disappears in older men, and as suggested by the common term "boxers' fractures" risk-taking, ghts and aggressive behaviours might lead to elevated numbers of young men incurring these injuries. However, our data suggests this behaviour might dissipate with age, or at least that a 1st incident decreases with age (since recurrent fractures are not reported in this paper)line. Conversely, distal phalanx fractures have a substantial predominance in men throughout the lifespan, except at the oldest age group. The fact that these fractures are less affected by age may re ect the nature of the risk activities. We hypothesize that gender-role differences in occupational and recreational exposures including outdoors work, equipment use, and manual labour, contribute to lifelong nger fractures risk exposures in men. In the oldest age group, the number of distal phalanx fractures is higher in women, but this may re ect the greater longevity of women and a larger cohort of living women who can be affected. Our hypotheses cannot be tested, given the limitations of health service data.
However, our hypotheses re ect the fact that sex and gender cannot easily be disarticulated when considering fracture epidemiology.
An important consideration when interpreting our data, particularly for the oldest age group is that the ratio between men and women in the population will vary over the life span and will become increasingly female predominant after the age of 80. In Canada, at age 80 the ratio of women to men is approximately 4:3, at 85 it is 5:3, at 90 it is 2:1 and at 95 it is 4:1 [26]. Thus, the higher numbers of women having fractures, particularly in fractures that are male predominant at younger ages likely re ects the larger volume of women in the population. Since we examine fracture volumes and proportions, not rates from the existing population these differences should not be misinterpreted as an increased risk for older women in comparison to men, when looking at the data for the oldest age group.
Females might have lifelong elevated fracture risks due to their bone geometry and composition [27]. However, at age 50 we see a dramatic increase in DRF and PHF even though population numbers are very similar for men and women. [26] We expected sex differences in these fractures due to osteoporosis/osteopenia preferentially affecting females [27]. These fragility fractures typically occur with low trauma, such a fall from level ground. The sexage patterns in our PHF and DRF data concurs with ndings from our mixed methods study where quantitative data from a cohort of more than 1400 patients with DRF [28] indicated that low trauma was more common in those over 45 years of age, and high trauma occurred most in the youngest (18-24-year-old) group. In the qualitative part of that study, we found the factors contributing to fractures included environment, risk taking behaviors, physical factors, and sports activities. Our data indicates that the excess of fractures in older female (50 + years) was also identi able in scaphoid fractures, radial head fractures, multiple fractures of the radius and ulna, and humeral shaft fractures which are less often considered as fragility fractures. Future studies should investigate the extent to which these less common upper extremity fractures predict osteoporosis, and future hip and spine fractures-and whether the de nitions of fragility fracture should be expanded.
One advantage of administrative data is the potential to study rare diagnoses. Monteggia fractures are an example, where we hoped to gain novel information on this rare but challenging fractures. Since only 50 cases were reported in our four-year cohort, longer cohorts would be needed for future investigations. We suspect that underreporting may partially explain these low numbers, since not all Monteggia fractures might be identi ed by the speci c code by nonspeciality physicians. Potentially these were classi ed as fracture dislocations rather than by their speci c name. Our data suggest that a coding validation study would be needed prior to conducting any administrative data analyses on this particular fracture type.
We found very low rates of tendon of nerve injury. The overall rate was in uenced by higher prevalence of nerve injury in nger and distal phalanx fractures; where the nerve injury is unlikely to cause permanent disablement. The higher proportion of 1.1% nerve injury seen in multiple fractures may relate to the severity of trauma and will likely have a more negative prognosis. Tendon injuries were also infrequently reported, and as expected were more common with nger fractures. Since no validation studies have been performed for nerve and tendon codes we cannot be con dent that these estimates re ect accurate attribution of the nature and severity of tendon and nerve injuries. It is possible that some nerve and tendon injuries are recognized/coded later when upper extremity specialist surgeons become involved in the case. This should be investigated in a future longitudinal validation study.
Age standardized rates of RA in Ontario, using ICES data, are reported to be between 0.49-0.9% [29]. The overall rate of RA in our cohort of 1.8% might suggest an elevated fracture risk, although we only examined fractures cases and could not calculate relative risks or age-adjusted rates. Olecranon fractures, DRF and fractures in multiple regions had greater proportions of RA diagnoses. The fact that the percentage of RA diagnoses were higher in women than men throughout all fractures is consistent with the sex distribution of RA [29]. Diabetes was present in 15.3% of the cohort; but elevated proportions were evident for olecranon fractures, DRF and PHF. ICES data have previously established that diabetes is associated with elevated risk of hip fractures in both men and women, even after controlling for age, greater comorbidity, being less likely to have had a BMD test, and more likely to be taking medications that increase risk of falling. [30] Another epidemiologic study found elevated fracture risk in patients with diabetes, for both upper and lower extremity fractures [31]. While overall the cohort with a 1st upper extremity fracture is relatively healthy, the prognosis of the subset with RA, diabetes, or elevated Charlson indices warrants further exploration, especially with respect to potential strategies to reduce secondary fractures.
One of the unique aspects of this study was the inclusion of indices of neighbourhood social deprivation which has been rarely studied as factor in fractures. Income had limited impact, with small excesses appearing in both upper and lower quartiles in certain fractures. PHF were the only type of fracture where a potentially relevant income effect was observed, which was an excess of fractures in the higher income double quartiles, for both sexes. Income may have been con ated with other more in uential social indices. Therefore, our conclusion is similar to that suggested by Karl et al [15] who concluded that SES, which they based on income, was not substantially related to fracture rates. They reported an excess of metacarpal and phalangeal fractures in their low-income group, which we did not nd. However, we did nd this excess in the dependency social indicator which encompasses employment. This supports our hypothesis that gendered employment exposures may partially explain "sex" differences in hand fractures. Conversely, while the prior study did not report that income affected PHF volumes; we did nd some effects. Age affects income indicators since older people with a PHF are more likely to be retired. The relationship between wealth, nancial independence and income is variable between people, methods for calculating the SES-income indicators, and across cohorts. We focused on volumes not rates, and de ned SES using multi-component indices which may have contributed to the minor differences in our ndings in comparison to those reported by Karl et al who uses rates and a single indicator of income [15].
With respect to deprivation, the only potentially relevant difference was for metacarpal fractures where the cohort excess in the lower double-quartile was 6.3%. When examining the sex-speci c proportions in the lowest quartile by fracture, this effect was present in males, but not in females. This index includes the impact of single parent families which might suggest that lack of parental supervision, or having a father in the home, which may have a larger impact on young men engaging in fracture-risk behaviours than it does for young women. This hypothesis has been empirically demonstrated in children [32]. However, our data can only highlight areas of differences at the neighbourhood level and while they may warrant exploration, causation cannot be inferred.
There were multiple differences that were potentially relevant with respect to ethnic concentration. DP, PHF, clavicle, scapula and multiple fractures were more prevalent in less ethnic neighbourhoods; whereas metacarpal, phalanx fractures, radial head and scaphoid had an excess in more ethnic neighbourhoods.
Ethnicity may affect fractures through multiple mechanisms including sex differences in genetic risks that affect the size and quality of bone, gender-based differences in behaviour norms that affect behaviours and life roles, and social differences that affect work and life exposures. African American and Hispanic men have been reported as having higher incidence and younger onset of hip fractures than white men [33], which is consistent with what we found for some upper extremity fractures, although not the classic fragility-subtypes.
Dependency was associated with some potentially relevant excesses in fractures, but the excess varied across fracture types. Metacarpal, phalangeal, radial head and scaphoid fractures had an excess in less dependent neighbourhoods. This may re ect the fact that employment contributes to this index and is a common source of injury in hand fractures. DRF was minimally affected by dependency which may re ect the fact that misfortune, bone fragility and frailty mechanisms are more important. Conversely, PHF had the highest dependency differential with almost a 20% excess in the upper double-quartile as compared to the lower. The may re ect a fragility/frailty pro le where larger numbers of older retired individuals experience these fractures. This is further informed by graph of the sex-speci c lowest quintile data which shows an excess preferentially in older women. The con ation of age and retirement in the dependency index, is consistent with PHF having the oldest mean age of the fracture subgroups.
The largest, and most unexpected, differences in social deprivation indices were observed with respect to residential instability which was observed as a global difference, and which was consistently manifested across most fracture subtypes. This index re ects residential home security, ownership and occupancy and is comprised of indicators of living alone, proportion of adults, number of people per dwelling, proportion who are single/divorced/widowed, and the proportion who have moved in the past 5 years. How these factors affect fracture risk is unclear and cannot be de ned in this observational cohort.
The potential mechanisms are many including precarious employment, higher rates of manual labour, less social support for assistance with tasks, risktaking, and environments with more fall hazards. Housing deprivation has been shown to relate to other health outcomes through complex mechanisms [34].
Social disadvantage can affect bone quality/strength to resist fractures due to poor nutrition during bone development, higher rates of smoking and less engagement in health promotion behaviours including bone-protective forms of exercise [35]. Social inequity is an important factor that affects health overall and that is not resolving. [36] Despite the limitations in neighbourhood social indices as an indicator of individual risk, we found substantial indication that social deprivation may affect fracture burden. This warrants more exploration of how social deprivation leads to fracture.

Limitations
Despite the large sample, this study has several limitations. Ontario is one province and may not represent other regions of Canada or other countries.
However, this data represents approximately 40% of the Canadian population and is one of the largest published upper extremity fractures cohort pro les.
Since administrative databases were designed for health system management and physician remuneration purposes, there are several inherent limitations to code coverage and validity. Not all fractures would present for care, and some fractures may be missed where imaging is not performed or able to detect early fracture (e.g. scaphoid fractures) or correctly interpreted/coded (e.g. Monteggia fractures). Social indicators were at the group level, not at the individual level, which limits precision.

Conclusions
The pro les of 1st incident fractures in patients without a prior history of upper extremity fracture as an adult is highly variable across different fracture types. Sex and lifespan greatly in uence these pro les. The patterns are suggestive of risk-taking behaviours, sex and gender mechanisms, physiological differences, sex and aging mechanisms that affect bone quality, comorbid health conditions, seasonal variations/environmental exposures and social roles/exposures as combining in uences on these pro les. Elicitation of mechanisms that increase risk across all of these domains and associated effective preventive strategies across all of these domains is likely to most effectively prevent secondary fractures. Variability of fracture pro les should be considered in case management. Major rst UE fractures in men and women Percentage of men/women with diabetes by fracture type Figure 24 Percentage of men/women with RA by fracture type